This week at COMPUTEX 2025 in Taipei, the global tech community gathers under the theme “AI Next” — a fitting lens through which to showcase the ongoing collaboration between VAST and NVIDIA. As AI innovation accelerates, the events this week highlight how infrastructure must evolve to match the scale and complexity of enterprise AI applications. VAST is here on the ground, demonstrating how our collaboration with NVIDIA is empowering new levels of performance, insight, and efficiency across real-world AI deployments.
A Long-Standing Partnership Built for the AI Era
VAST and NVIDIA have spent years developing solutions at the intersection of accelerated computing and modern data management. From our earliest collaborations pioneering VAST’s disaggregated, shared-everything (DASE) architecture with NVIDIA DGX systems, to our joint collaboration in network and storage offload technologies as well as composable AI infrastructure, this partnership continues to deliver innovation that enables enterprises to unlock the full potential of their data. Today, this partnership is more essential than ever - delivering deeply integrated solutions across software and hardware to accelerate enterprise AI readiness and impact.
Unlocking New Insight from Enterprise Data with NVIDIA AI Software
At the heart of the VAST and NVIDIA collaboration is a shared vision: to make all enterprise data actionable for AI. Today, VAST announced how we’re building the data foundation for the agentic enterprise together with NVIDIA, to deliver the world’s first data and computing platform purpose-built for enterprise-grade, multi-agent AI systems. This isn’t a reference architecture or a demo — it’s a production-ready foundation built to help enterprises unlock new levels of automation, reasoning, and operational efficiency by harnessing AI on both historical and real-time data from across their internal systems, files, and databases.
By leveraging the NVIDIA AI-Q blueprint and a powerful suite of NVIDIA software technologies — from NVIDIA NeMo and NVIDIA NIM microservices to NVIDIA Dynamo, NVIDIA Agent Intelligence toolkit, and more — alongside the VAST InsightEngine and our VUA acceleration layer, this platform empowers enterprises to deploy AI agent systems capable of reasoning over enterprise data, to deliver faster, smarter outcomes. It’s a new kind of enterprise AI stack, built for the era of AI agents — and VAST is proud to be leading the way.
This follows previous examples shown at NVIDIA GTC in March: VAST recently demonstrated an enhanced implementation of NVIDIA’s multimodal PDF extraction blueprint built on NIM and NeMo Retriever, which delivered 15× faster ingestion and 50% higher accuracy in extracting text, charts, and tables from documents. Uniquely, VAST embedded unified RBAC permission enforcement throughout the pipeline — ensuring secure, role-aware retrieval and summarization of sensitive enterprise content.
VAST also worked with NVIDIA to showcase the NVIDIA AI Blueprint for video search and summarization (VSS), demonstrating how AI can semantically index and retrieve insights from VAST video archives — like the National Hockey League’s entire media collection. With sub-second semantic search performance across petabytes of content, the demo illustrated how VAST enables fast, secure AI-powered discovery — with built-in governance and data controls that enterprises demand.
Ushering in a New AI Infrastructure Era with the AI Data Platform
At NVIDIA GTC 2025, NVIDIA introduced the NVIDIA AI Data Platform — a new blueprint and reference architecture that positions infrastructure not as a passive layer, but as an active enabler for modern AI agents and large-scale model operations. It’s a vision that directly aligns with VAST’s founding mission.
By unifying accelerated computing, low-latency networking, and flash-optimized data services, VAST has fundamentally redefined how infrastructure should be built for AI-native workloads. This shift isn’t just semantic — it’s structural. Storage is no longer a static vault where data waits to be summoned. It’s a dynamic, intelligent system that embeds context, extracts meaning, and responds in real time — a living substrate that thinks alongside compute.
VAST hasn’t just anticipated this future — we’ve been actively building it. The VAST Data Platform’s agentic-ready architecture already delivers a global, context-aware namespace that allows AI systems to ingest, interpret, and act on data instantly, without the delays of staging, transformation, or middleware. It’s a foundation built to make data immediately usable and meaningful at AI speed.
AI isn’t just reshaping computation, it’s redefining the nature of information itself. In this new AI factory model, where systems don’t just compute but generate, reason, and adapt, data isn’t an accessory … it’s the fuel. Structured and unstructured. Historical and real-time. Not just retrievable, but immediately understood.
The VAST InsightEngine extends this capability even further, adding real-time indexing, vector search, and multi-modal reasoning atop the core platform. It empowers AI agents to reason over petabyte-scale enterprise data sets with the speed, precision, and governance enterprises demand.
This tight alignment with NVIDIA AI Data Platform reflects years of joint development and shared ambition - positioning VAST as a foundational technology partner for enterprises building AI factories at scale, whether on-premises, in the cloud, or across hybrid environments.
Pushing the Boundaries of Storage Infrastructure with NVIDIA BlueField DPUs
At COMPUTEX, VAST is also announcing support for its next-generation Ceres platform — a highly efficient NVMe enclosure that forgoes traditional x86 controllers in favor of NVIDIA BlueField DPUs. The new version of Ceres (from our friends at AIC) offloads data services and IO processing directly to new NVIDIA BlueField-3 DPUs, enabling VAST systems to further improve storage throughput, power efficiency, and scalability. A single 1U Ceres node delivers over 60 GB/s of sustained bandwidth - making it ideal for the high-throughput demands of AI workloads.
Ceres is already qualified a the foundational storage layer for NVIDIA DGX SuperPOD deployments — reinforcing the tight integration of VAST hardware with NVIDIA solutions. With BlueField-3 DPUs now in production, VAST systems can natively execute data-plane services at wire speed, bringing compute, networking, and storage into closer alignment than ever before. This evolution sets the stage for new classes of scale-out AI systems, where infrastructure grows linearly alongside the size and ambition of AI models.
The Road Ahead: Built Together for What’s Next
As AI transforms how we work, live, and compute, infrastructure must evolve just as quickly. VAST with NVIDIA has consistently delivered innovation where it matters most:
Software integrations that make enterprise data AI-ready
Architectural breakthroughs that eliminate bottlenecks
Hardware advancements that expand the performance frontier
This is a blueprint for the future of AI infrastructure. And it’s happening now.
If AI is the next great industrial revolution, VAST and NVIDIA are building the factories.
In next week's VAST/NVIDIA webinar you can learn more about building intelligent, multimodal AI agents that reason in real time, powered by the combined force of the VAST Data Platform and NVIDIA’s AI-Q. Register here.
